Claude Multi

Run multiple Claude accounts in one window. Auto-switch when one hits its limit.

🌐 claude-multi website · ⬇ Download Free for Windows · macOS / Linux · Quick Start

⭐ If this saves you from a usage wall — please star the repo! It helps others find Claude Multi when they search for it.

Claude Multi running on Windows — multiple accounts, live cooldown timers, streaming chat

What is Claude Multi?

Claude Multi is a free, open-source desktop app for Windows, macOS, and Linux that lets you run Claude Code with multiple Claude accounts — all in one window.

The core problem it solves: Claude's 5-hour usage limit stops your work cold. Claude Multi watches for that limit, instantly switches to your next available account, copies your conversation over, and re-issues the interrupted instruction automatically — so your task keeps running without you lifting a finger.

Popular use cases:

• 🔄 You have 2–5 Claude Pro/Max accounts and want to use them without switching browsers
• 🤖 You want Claude to work 24/7 on a long task while you sleep (Auto-loop)
• 💬 You want to run parallel chats on multiple projects simultaneously
• 📊 You want to see exactly what is consuming your Claude context and quota
• 🔋 You want to stretch your Claude subscription as far as possible

No API key required. Works with your normal Claude subscription login. 100% local — no telemetry, no servers, your data never leaves your machine.

🆕 What's New in v1.30.0

• 🧠 Claude Opus 5 + the full model line-up — Fable 5, Opus 5, Opus 4.8/4.7/4.6, Sonnet 5/4.6, and Haiku 4.5, all in the model picker.
• ⚡ Five effort levels — Low, Medium, High, Ultra think (xhigh), and Max, matching Anthropic's effort parameter exactly. Dial reasoning depth per chat.
• 🎤 Offline voice-to-text — dictate straight into the composer with a bundled ~40MB speech model. Fully offline, no API key, no cloud, 100% private.
• 🎯 Redesigned account switcher — bigger targets, clean stacked layout, a clear "Active" badge, and one-key switching (Ctrl/Cmd + 1…9).
• 🧵 Background chats stay live — switch away from a generating chat and back; its streaming response is preserved.

See the full release notes.

Key Features

Download

Free. No account. No subscription. Just download and run.

Windows SmartScreen warning: The app is not code-signed yet. Click More info → Run anyway. The full source is open — read it here.

Prerequisites: Node.js 18+ and Claude Code CLI must be installed on your machine.

Quick Start

Option A — Installer (2 minutes)

1. Download the installer from Releases
2. Run it, follow the steps, open Claude Multi
3. Click New chat → pick your project folder
4. Open the account switcher (bottom-left) → Add account → Log in
5. Sign in when the browser opens — done ✓

Option B — Run from source

git clone https://github.com/Chamanrajragu/claude-multi.git
cd claude-multi
npm install
npm start

Add more accounts:
Account switcher (bottom-left) → Add account → Log in → repeat for each Claude account.
Claude Multi remembers which account to use per project folder.

Full Feature List

Multiple Claude Accounts — Seamless Switching

• Add up to 20 Claude accounts, each fully isolated via CLAUDE_CONFIG_DIR
• Auto-switch on usage limit — re-issues your last instruction on the next account
• One-click manual switch with Ctrl/Cmd + 1…9
• Conversation transcript carried over automatically — no copy-pasting
• Per-project account memory — each folder remembers its preferred account
• Activity dashboard: every account's status, cooldown timer, and token usage
• Manual cooldown clear if needed

24/7 Auto-Loop — Run Claude While You Sleep

• Set a prompt once — Claude Multi runs it on all accounts in rotation, forever
• Fresh chat per account per round — context stays small and cheap
• Live per-account status table with round status and cooldown countdown
• Round counter, total sends, and wait-reason display
• System tray integration — see auto-loop status and stop it without opening the app

Chat Experience

• Parallel chats — multiple projects open at once, each independent
• Real-time streamed markdown rendering
• Syntax-highlighted code blocks with language labels and one-click copy
• Red/green diff view for every file Claude edits
• Collapsible tool cards (📖 Read · ✏️ Edit · ⌨️ Bash · 🔍 Search · 🌐 Web)
• Allow / Deny permission cards before Claude modifies any file
• Thinking blocks in collapsible panels
• Edit and resend any previous message
• Retry to regenerate the last response
• /compact command to summarize the chat instantly

Composer

• Slash commands — type / for /new, /compact, /model, /switch, /copy and more, without leaving the keyboard
• @file autocomplete — type @ to fuzzy-search any file in your project
• Voice to text (offline) — mic button dictates straight into the composer using a bundled ~40MB speech model; no internet or API key required
• Prompt templates — save and reuse your best prompts
• Paste images directly — screenshots, browser images, any format
• Drag and drop files onto the window
• Character counter with warning at 8K characters
• Prompt history — Up/Down arrow to recall previous messages
• Per-chat draft saving — unsent text preserved when switching chats

Token and Quota Management

• Context inspector (Ctrl/Cmd + Shift + K) — every item with token counts
• Live token meter — percentage used and tokens remaining, always visible
• Auto-compact — summarizes before the context overflows
• Long-chat warning — nudges you to compact or start fresh
• Budget caps — stop a turn after N steps or $X spend
• Disable unused tools — subagents, web search, fetch, todos, notebooks
• Every current model — Fable 5, Opus 5, Opus 4.8/4.7/4.6, Sonnet 5/4.6, Haiku 4.5
• Five effort levels — Low, Medium, High, Ultra think (xhigh), Max — set per chat
• Per-chat model and effort — Opus for hard tasks, Haiku for quick edits
• Least-used account routing — starts new tasks on the account with most quota

Navigation and UI

• Command palette (Ctrl/Cmd + K) — every action and chat, instantly
• Workspaces — save a project folder paired with an account, then reopen the combo in one click
• Global summon shortcut — raise the window from any app, and press again to hide it
• Full-text search across all chats (Ctrl/Cmd + F)
• Find in chat (Ctrl/Cmd + Shift + F)
• Drag-to-reorder chats in the sidebar
• Pin chats to keep important ones at the top
• Alt + ↑/↓ to move between chats
• Recent projects — right-click New Chat for instant folder switching
• Resizable sidebar
• Zoom (Ctrl/Cmd + = / − / 0)

Reliability

• Session auto-restart — expired sessions detected and restarted with a toast notification
• Undo delete — a deleted chat can be restored from the toast before it is discarded
• Background chat saving — off-screen chats save correctly
• Desktop notifications when a chat finishes or needs your approval
• Export to Markdown — single chat or all chats at once, or copy one straight to the clipboard
• Save any code block to a file, with the extension picked from the fence language
• Update check — tells you when a newer release is published, and can be switched off
• Backup and restore — export/import your settings and account list

Screenshots

Main chat	Auto-loop running 24/7

Usage dashboard	What's using my context

Account switcher	Usage limit — carry the chat over

Command palette	Token settings

Model picker	Light theme

Keyboard Shortcuts

How It Works

Each account gets its own ~/.claude-accounts/<id>/ directory — the officially documented CLAUDE_CONFIG_DIR mechanism — so logins, sessions, and settings never collide.

Chats run through the official Claude Agent SDK (@anthropic-ai/claude-agent-sdk) using your existing subscription login. The first-time sign-in opens claude /login in a small terminal window (OAuth requires an interactive session). After that, everything runs silently in the background.

The auto-switch flow, step by step:

Your task is running on Account A
       ↓
Account A hits usage limit
       ↓
Claude Multi detects the limit error in the stream
       ↓
Account A stamped with cooldown (reset time from server response)
       ↓
Account B is free → transcript copied → session resumed with --resume
       ↓
Your interrupted instruction re-issued on Account B
       ↓
Task continues — you see a brief "switched to Account B" toast

Making Your Claude Quota Last Longer

Every turn re-sends the full context — conversation history + CLAUDE.md + tool descriptions. Claude Multi gives you fine-grained control to slash token waste:

Account Safety and Terms

Claude Multi is built for one person using their own accounts — not sharing, pooling, or reselling access.

Each account authenticates with its own normal subscription login and stays within its own quota. The CLAUDE_CONFIG_DIR mechanism is officially documented by Anthropic.

Important caveat: Since February 2026, Anthropic's terms of service restrict using Pro/Max subscription credentials in third-party tools (including the Agent SDK). Claude Multi does exactly that. Account restrictions can happen without warning.

The honest summary: this is use-at-your-own-risk. If you want zero risk, use Claude Code or claude.ai directly. Read Anthropic's terms before deciding.

Privacy — What Claude Multi Does NOT Do

Everything is verified in the open source code:

• ❌ No telemetry
• ❌ No analytics
• ❌ No crash reporting
• ❌ No update pings
• ❌ No outbound network requests from the app itself
• ✅ Conversations go only between your local claude process and Anthropic's servers
• ✅ OAuth tokens are written by Claude Code locally — Claude Multi never reads or transmits them
• ✅ All data stays in Electron's local userData folder on your machine

Build from Source

# Install dependencies
npm install

# Development mode (live reload)
npm start

# Build distributable for your current platform
npm run dist

# Build for specific platforms
npm run dist:win      # Windows installer + portable .exe
npm run dist:mac      # macOS .dmg (must run on macOS)
npm run dist:linux    # Linux AppImage

Creating a release — tag it and GitHub Actions builds everything automatically:

git tag v1.X.0
git push --follow-tags

Tests

npm test

Covers: usage-limit detection, reset-time parsing, account selection logic, cooldown math, and the account store — including thousands of fuzzed inputs.

FAQ

How do I use Claude with multiple accounts in one application?
Install Claude Multi, click Add account once per Claude subscription you own, and sign in to each. Every account gets its own isolated CLAUDE_CONFIG_DIR, so the logins never collide. Then just chat — when one account hits its limit, Claude Multi moves the conversation to a free account and re-sends the interrupted instruction. Full walkthrough: Quick Start.

What happens when Claude hits the 5-hour usage limit?
Claude Multi reads the limit message, parks that account with a live countdown to its real reset time, then switches to an account that is free right now — copying the transcript across and re-issuing the instruction that got cut off. Your task continues instead of starting over.

Can I run two Claude accounts at the same time?
Yes. Each chat carries its own folder, its own account and its own session, so you can run several accounts in parallel on different projects simultaneously. Background chats keep working while you look at another one.

Do I need a Claude API key?
No. Claude Multi uses your normal Claude Pro / Max / Team subscription login — the same one you use with Claude Code. No API key, no extra cost.

Is Claude Multi free?
100% free and open source (MIT license). No paid tier, no account required, no ads, no upsells.

What operating systems does it support?
Windows, macOS, and Linux.

Is it against Anthropic's terms of service?
It is not account-sharing. But Anthropic's Feb 2026 terms restrict subscription logins in third-party tools. Use at your own risk — see Account Safety.

How many Claude accounts can I add?
Up to 20 accounts.

Why is Claude using so many tokens?
It is almost always the context window, not your prompts. Open the Context Inspector (Ctrl/Cmd + Shift + K) to see a breakdown. See Making Your Claude Quota Last Longer.

Can Claude Multi run tasks while I sleep?
Yes — that is what the Auto-loop feature is for. Set your prompt once and it runs on all accounts in rotation, waiting through resets, indefinitely.

Does it work with Claude Code?
Yes — Claude Multi runs Claude Code in headless mode via the official Agent SDK. It is a GUI wrapper and multi-account manager for Claude Code.

Can I use this if I only have one Claude account?
Yes. Single-account mode works fine — you still get parallel chats, the context inspector, @file autocomplete, the composer features, and everything else.

Related Tools and Alternatives

• Claude Code — Anthropic's official CLI (Claude Multi is built on top of it)
• claude.ai — the official web interface
• Looking for API-based multi-account? Claude Multi uses subscription logins, not API keys

Author

Built and maintained by Chaman Raj — github.com/Chamanrajragu

Found a bug? Open an issue.
Have an idea? Start a discussion.

If Claude Multi saved you from a usage wall — a ⭐ on this repo helps thousands of other people find it when they search for a solution. It takes 2 seconds and means a lot.

License

MIT © 2026 Chaman Raj

Claude Multi is not affiliated with or endorsed by Anthropic. "Claude" and "Claude Code" are trademarks of Anthropic, PBC.